[
https://issues.apache.org/jira/browse/OPENNLP-715?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14353632#comment-14353632
]
Rodrigo Agerri commented on OPENNLP-715:
----------------------------------------
Refactorings done. The current WordClusterDictionary class reads Word2vec and
Clark clusters. There are the most common, as other toolkits (e.g., producing
LDA clusters) also produce this format.
I think that the prefix consuming memory could be an issue if many cluster
lexicons are used, but so far in the literature it is quite rare for a system
to use more than two or three different cluster lexicons. Still, if that is
considered to be an issue, I could create a new ClarkClusterDictionary and
ClarkFeatureGenerator and decouple the functionality currently in
WordClusterDictionary for each type of cluster.
Either way I can do it by tomorrow.
> Clark clusters NameFinder features
> ----------------------------------
>
> Key: OPENNLP-715
> URL: https://issues.apache.org/jira/browse/OPENNLP-715
> Project: OpenNLP
> Issue Type: New Feature
> Components: Name Finder
> Affects Versions: 1.6.0
> Reporter: Rodrigo Agerri
> Assignee: Rodrigo Agerri
> Priority: Minor
> Fix For: 1.6.0
>
>
> Add token based features from Clark clusters (Clark 2003). This feature is
> actually the same as the one implemented in the WordClusterFeatureGenerator,
> but we should somehow make them separate (perhaps implementing a dynamic
> prefix id for each one, as in the dictionary features) as it has been shown
> that the combination of these clustering-based features improve results.
> Clark clusters can be generated using this tool:
> https://github.com/ninjin/clark_pos_induction
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
